The U.S. Department of Energy has launched the Genesis Open Models Initiative, a new program to build open-weight AI foundation models for scientific discovery — the first U.S. government-backed effort of its kind.
Announced on August 7 as part of DOE's broader Genesis Mission, the initiative aims to give researchers, national laboratories and industry open, transparent and adaptable models for domains including materials discovery, energy systems, earth-system modeling, fusion, biology and high-energy physics. The first model, Genesis-Science-1, is being developed in partnership with Arcee AI.
'By releasing models with open weights, DOE seeks to galvanize the scientific and AI communities around shared infrastructure for science,' the department said, citing open science, reproducibility and responsible AI as guiding principles.
DOE is now soliciting input from commercial, academic and research institutions in three categories: open-weight base models; pretraining contributions such as curated datasets, benchmarks and specialized scientific corpora; and fine-tuning efforts to shape domain-specific versions — for example lab assistants, simulation surrogates and scientific copilots. Applications for the first pretraining window close August 14 and for fine-tuning August 25, with rolling deadlines every three months afterward.
The initiative marks a milestone in AI policy: a major government entering the open-weight arena as a direct participant and builder, rather than only a funder or regulator — and betting that transparent, shared scientific AI models can be a public good.
